Skip to main content
Korea Says Nvidia Keeps 260,000-GPU Pledge After Jensen Huang Quits; Only 10,000 Confirmed in Use - Tech Times

Korea Says Nvidia Keeps 260,000-GPU Pledge After Jensen Huang Quits; Only 10,000 Confirmed in Use - Tech Times

NVIDIA GPUs
16 lis 2026, 05:47Hardware News
Oryginał
3 min read

South Korea's science ministry said on Nov. 13 that Nvidia reaffirmed its 260,000-GPU pledge one day after Jensen Huang resigned as CEO, though only 10,000 of the chips are confirmed to be in use.

The resignation of Nvidia CEO Jensen Huang has not halted the massive state-backed push for artificial intelligence in South Korea. On Nov. 13, Deputy Prime Minister and Science Minister Bae Kyung-hoon announced that Nvidia had reaffirmed its commitment to a 260,000-GPU supply plan at the corporate level.

Bae made the announcement in a Facebook post titled 'Jensen Huang's resignation, but cooperation continues,' stating that both sides agreed to keep existing projects running without disruption.

The reassurance came after a meeting between Korean officials and Nvidia to discuss pending issues, though the tech giant itself has not issued a public statement on the pledge.

Nvidia said the board expects 'operations, customer commitments and product roadmap' to continue without interruption.

Gap Between Secured and Installed Chips

Despite the high-level pledge, the practical situation for South Korean researchers and startups is less clear. The ministry's figures show a stark difference between the number of chips 'secured' and those actually in service.

Of the roughly 35,000 GPUs the government buys directly, about 21,000 are secured, or around 60%, in B200-equivalents. The remaining 14,000 are expected by early 2027.

The only confirmed figure for chips currently in use is 10,000, as stated by Vice Minister Ryu Jemyung on Nov. 9. This number may or may not be part of the 21,000 secured figure.

The ministry uses 'B200-equivalent' as a normalized unit, which allows chips from different generations to be counted together based on peak compute. A B200 delivers roughly 9 petaFLOPS at 8-bit precision.

Korea's Sovereign AI Ambitions

The 260,000-GPU package is central to South Korea's strategy to become a top-three global AI power, a goal that has drawn both praise and caution from local experts.

Bae, who previously led LG AI Research, frames the Nvidia relationship as a foundation for physical AI, joint research, and talent development. The cooperation includes linking Korean AI chips to Nvidia systems through NVLink.

However, Lee Seung-hyun of the Digital Platform Government Committee warned that relying on Nvidia's ecosystem for software and model training could leave Korea unable to design its own 'nervous system' over the long term.

A recent announcement involving Nvidia, Rebellions, and SK Telecom showed the limits of this independence. The plan to connect Rebellions' R200 chip to Nvidia systems via NVLink Fusion requires adopters to still include Nvidia silicon in each node.

Uncertainty Ahead for AI Teams

For Korean AI teams relying on state-funded compute, the key question is not just the total number of chips, but their physical location and generation.

Three markers will determine if the pledge holds after Nvidia reports its third-quarter fiscal 2027 results on Nov. 17: the delivery pace of the remaining 14,000 chips, the product mix (specifically whether Vera Rubin parts are prioritized), and the status of corporate tranches for Samsung, SK, Hyundai, and Naver.

The resignation itself was driven by a board inquiry into Huang's relationship with a former executive at a Singapore-based chip company, allegations that Beijing has called groundless.

As the interim CEO, Colette Kress faces the immediate task of maintaining operations and customer commitments during this transition.

Z.ai
Article written by AI
Z.ai GLM-4.7-FlashFree
Free AI articles150 of 150
No limit after signing in with Telegram